A newly published Viewpoint in BioScience argues that the continuing dispute over the origins of SARS-CoV-2 should be addressed through a more rigorous and transparent scientific process. The article, written by Alan B. Franklin, formerly of the US Department of Agriculture’s National Wildlife Research Center, does not endorse a particular explanation for the emergence of the virus. Instead, it examines how scientists should compare competing hypotheses when evidence is incomplete, politically sensitive and unevenly distributed. Franklin focuses on two broad possibilities that have dominated public debate: an unintentional release associated with research or laboratory activity at the Wuhan Institute of Virology, and a zoonotic emergence involving infected animals connected to the Huanan Seafood Wholesale Market or related wildlife-trade networks in Wuhan. His central argument is methodological rather than accusatory: the origin question requires structured inference, explicit comparison of alternatives and independent evaluation of evidence.
Franklin’s proposed framework draws on the “method of multiple working hypotheses,” an approach described by the American geologist Thomas Chrowder Chamberlin in an 1890 paper. Chamberlin contrasted this method with what he called the “method of the ruling theory,” in which investigators become attached to one explanation and then interpret new observations primarily as support for that preferred account. Under a ruling-theory approach, evidence that appears inconsistent with the favored hypothesis may be minimized, reinterpreted or dismissed, while observations that fit it receive disproportionate attention. Multiple working hypotheses are intended to reduce that tendency by requiring investigators to formulate several plausible explanations at the outset and to assess each against the same body of evidence. Franklin argues that the origins of COVID-19 are especially suitable for this approach because no single line of evidence currently resolves the question.
The concept is closely related to the principle of strong inference, a term associated with the physicist and philosopher John R. Platt. Strong inference involves establishing alternative hypotheses, deriving predictions from each one and then seeking observations or experiments that can discriminate among them. The objective is not simply to accumulate information but to identify evidence with different expected outcomes under competing explanations. For a virus-origin investigation, that could mean asking whether a particular genomic feature, epidemiological pattern, laboratory record, wildlife sample or supply-chain connection would be more likely under a natural-spillover hypothesis than under a laboratory-associated hypothesis. A useful analysis must also identify what findings would weaken each explanation. Franklin emphasizes that hypotheses should be evaluated symmetrically, rather than judged according to different standards of proof.
The article highlights a persistent difficulty in origin investigations: negative evidence is often difficult to interpret. The failure to find a precursor virus in wildlife, for example, does not demonstrate that such a virus never existed. Sampling is limited by geography, season, animal behavior, species availability and the time elapsed between an outbreak and the collection of specimens. Similarly, the absence of a documented laboratory incident does not establish that no accident occurred, but neither does the existence of laboratory work involving related coronaviruses demonstrate that SARS-CoV-2 emerged from that work. Evidence may be missing because records were not preserved, samples were destroyed, animals were moved through undocumented channels or investigators did not know which clues would later become important. Franklin therefore cautions against treating the volume of available information as a measure of evidentiary strength. A large dataset can still be weak if it is biased, indirect or incapable of distinguishing among hypotheses.
SARS-CoV-2 provides a particularly complex case because different forms of evidence address different stages of emergence. Viral genomes can reveal relationships among strains, estimate evolutionary divergence and identify patterns of adaptation, but they do not by themselves specify the location or mechanism of transmission into humans. Epidemiological data can show where the earliest recognized cases occurred and whether infections clustered around a particular site, yet early case records may be incomplete and may reflect where patients sought care rather than where exposure occurred. Environmental sampling can detect viral RNA in market stalls or drainage systems, but such material may originate from infected people rather than from an infected animal. Laboratory records, biosafety procedures and inventories may clarify what research was conducted, although their interpretation depends on access, completeness and independent verification. The article’s framework would require each evidence stream to be assessed for reliability, relevance and discriminatory power.
Franklin points to natural-resource management as a field that has developed practical tools for making decisions under uncertainty. Wildlife biologists and conservation agencies frequently confront situations in which several explanations can account for the same observation, such as the decline of a species, the spread of a pathogen or the failure of a habitat-restoration program. In these settings, investigators may agree in advance on criteria for evaluating evidence, assign relative support to competing models and update conclusions as new data become available. Such procedures are not intended to produce artificial numerical certainty. Rather, they make assumptions visible and prevent participants from changing standards midway through an investigation. Applied to SARS-CoV-2, a comparable process could define what kinds of findings would constitute strong, moderate or weak support for each origin hypothesis before researchers examine the evidence in detail.
The Viewpoint also calls for a prominent role for scientific societies and other independent institutions. Franklin argues that a broad consortium could bring together virologists, epidemiologists, evolutionary biologists, wildlife experts, biosafety specialists, statisticians, social scientists and scholars of scientific reasoning. A diverse group would not eliminate disagreement, but it could improve the quality of the debate by making methodological decisions explicit and ensuring that relevant expertise is not concentrated in a small number of individuals or organizations. The process would also benefit from conflict-of-interest disclosures, access to underlying data and records, independent replication of analyses and clear separation between scientific assessment and political messaging. Franklin maintains that institutions capable of convening such groups may be better positioned than individual researchers or political actors to communicate uncertainty without appearing to advocate for a predetermined conclusion.
The proposal reflects a broader concern about how politically charged scientific questions are communicated to the public. In the SARS-CoV-2 origin debate, discussions of laboratory safety, wildlife trade, international transparency and national responsibility have often been intertwined with claims about virology and epidemiology. This has made it difficult for non-specialists to distinguish direct evidence from inference, and testable hypotheses from speculation. Franklin does not suggest that every explanation deserves equal weight merely because it has been proposed. Multiple working hypotheses require plausibility, testable implications and willingness to discard ideas that conflict with reliable observations. At the same time, the method discourages premature closure, particularly when the available record is incomplete. Its purpose is to calibrate confidence, not to manufacture consensus.
The article concludes that a stronger analytical process could help restore confidence in scientific institutions while improving the search for SARS-CoV-2’s origins. Franklin states that his objective is not to establish whether the pandemic began through natural spillover, an accidental laboratory-associated event or another pathway, but to promote a fairer method for distinguishing among those possibilities. The approach would begin by defining competing hypotheses, identifying predictions and agreeing on evaluation standards before reviewing the full evidence. It would then require continuing updates as new viral sequences, animal samples, epidemiological records or laboratory documentation become available. For a question that remains scientifically unresolved and socially divisive, Franklin argues, transparent reasoning may be as important as any individual discovery. A process based on multiple working hypotheses and strong inference cannot guarantee a definitive answer, but it can make conclusions more robust, accountable and scientifically credible.
Subject of Research: Scientific methods for evaluating competing hypotheses about the origins of SARS-CoV-2 and the COVID-19 pandemic
Article Title: Multiple working hypotheses, strong inference, and the origins of the COVID-19 pandemic
Web References: https://doi.org/10.1093/biosci/biag037
References: Alan B. Franklin, “Multiple working hypotheses, strong inference, and the origins of the COVID-19 pandemic,” BioScience; Thomas Chrowder Chamberlin’s 1890 paper describing the method of multiple working hypotheses
Keywords: SARS-CoV-2, COVID-19, viral origins, zoonotic spillover, laboratory-associated emergence, virology, epidemiology, strong inference, multiple working hypotheses, scientific methodology, pandemic origins
